Dermatomyositis (DM) is a rare autoimmune disease characterized by chronic muscle inflammation, weakness, and skin rashes. Cutaneous Dermatomyositis (CDM), the skin manifestation of the disease, typically presents as purple or red rashes on the eyelids, joints, knuckles, and other areas; while there is no cure, treatment can alleviate symptoms, and monitoring disease progression is crucial. This study introduces a novel image-based approach for assessing CDM severity, aiming to create an objective, predictive model based on dermatological images, with expert assessments of the Cutaneous Dermatomyositis Activity Score as the gold standard. Through our collaboration with clinicians at the University of Pittsburgh Medical Center, we analyze a dataset of high-resolution 3D in-clinic hand images from DM patients. Key clinical features, including the extent, intensity and texture of the rash, are analyzed alongside CNN-based image features, enabling a comprehensive assessment of disease severity. We evaluate multiple state-of-the-art image classification models, fine-tuning them on our dataset to optimize performance. Our approach includes utilizing semantic image segmentation to accurately highlight regions of interest, with significant improvements achieved through this integration. Our study lays the groundwork for the use of patient-taken images for remote monitoring, demonstrating the potential for patients to track their condition at home. By combining clinical insights with advanced image analysis, this work contributes to improved automated assessment of CDM and better monitoring of disease progression.

67 pages

Thesis Committee:
Artur W. Dubrawski (Chair)
Bhiksha Raj

Srinivasan Seshan, Head, Computer Science Department
Martial Hebert, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu